New GAT 2.2 coming soon!

G.A.T. Engine general discussion
User avatar
lottoarchitect
Site Admin
Posts: 1534
Joined: Tue Jan 15, 2008 5:03 pm
Location: Greece
Contact:

New GAT 2.2 coming soon!

Post by lottoarchitect » Sun Feb 05, 2012 12:43 pm

Well, a break-through in the GATs operation is coming the next weeks. Current GAT has 2 major "flaws". The need to wait for the whole process to complete in order to make a selection and huge memory demands to run it at its limits. Also it has a limited maximum GATs allowance because it has to store all the GATs computed, although most are not the best that can be found therefore no real candidates to pick from. All these will become history now! The new version will offer:
- Virtually unlimited GATs (up to 1 billion).
- Instant minimum hits performance. The more you let the engine run (explore more GAT tables), better overall hits will emerge. It is up to the user to decide how long will let the engine run for.
- The process can be stopped at any time, so to utilize the best GATs that have been found till that point.
- Drastically reduced memory demands. A typical run will need at most 200MB of memory.
- No more GUI delays.

A typical overnight run will be able to compute many millions of GAT tables which will contain more hits than the maximum possible that can achieved with the current engine (due to maximum possible GATs that can be explored). It is possible to get predictions right away (minimum possible hits), no matter how many tested draws or statistical data are used.

The following is an example of potential increase in hits production when tested at a 1/20 field (50 tested draws) (explored 20 million tables vs 750000 maximum of current GAT - run time was approx. 6 hours):

Pick 1: 15 (new version) vs 13 (current version) hits
Pick 2: 21 vs 17
Pick 3: 24 vs 21
Pick 4: 27 vs 25
Pick 5: 30 vs 30
Pick 6: 36 vs 31
Pick 7: 39 vs 34
Pick 8: 41 vs 39
Pick 9: 42 vs 42
Pick 10:44 vs 42

Another example of the "lucky star" field 2/11 of the euromillions game (40 tested draws, 20 stat. data. 60M tables (new) vs 500000 of current GAT):

Pick 1: 23 (new version) vs 20 (current version) hits (1 correct)
Pick 2: 31 vs 29 (1+ correct), 9 vs 8 (2 correct)
Pick 3: 36 vs 34 (1+ correct), 14 vs 11 (2 correct)
Pick 4: 40 vs 39 (1+ correct), 18 vs 16 (2 correct)
Pick 5: 40 vs 40 (1+ correct), 23 vs 20 (2 correct)
Pick 6: 40 vs 40 (1+ correct), 30 vs 26 (2 correct)
Pick 7: 40 vs 40 (1+ correct), 33 vs 31 (2 correct)

cheers
lottoarchitect

elmoo168
Casual
Casual
Posts: 16
Joined: Wed Dec 21, 2011 4:55 pm

Re: New GAT 2.2 coming soon!

Post by elmoo168 » Tue Feb 07, 2012 4:15 am

I can't wait!!! Can you tell us when we can expect to play with the new release???

Also I never could figure out exactly what a GAT table is. Is each GAT table a different number selector algorithm or is it the same combination of numbers that's tested across the actual draw results?

Thanks!

User avatar
lottoarchitect
Site Admin
Posts: 1534
Joined: Tue Jan 15, 2008 5:03 pm
Location: Greece
Contact:

Re: New GAT 2.2 coming soon!

Post by lottoarchitect » Tue Feb 07, 2012 9:11 am

Hi elmoo168,

expect this release in about 2 weeks from now.
As for what is a GAT table, I can't answer that easily without giving much details of the inner workings but I'll give it a shoot. Lotto draws have reduced randomness, which means to most people, including me, they seem to be completely random and therefore impossible to make a prediction for the next draw. Reduced randomness however means short-term biased (NOT RIGGED) which means a set of few past draws share a common characteristic. I can't really describe how such a characteristic looks like but think of it as points (which I call attractors/repellers - the signature) which have the distinct effect of telling how we move from one draw to the next over that small set of draws analysed to extract this signature. Having knowledge on that signature, we can estimate how the next draw might be like. This is what a GAT table do; estimate signatures and make a prediction. The signature in practice try to "visualize" the inherent randomness of the draws. The very first GAT #1 computed has a very basic such analysis of randomness. It passes its own understanding of randomness to GAT#2 which refines this randomness further. This goes on and on for as long as we want it to. So each GAT X actually represents a set of signatures as evolved and influenced by all GATs #1-#X-1. So in order to compute GAT #1000, we need the information of GATs 1-999. During that process, GATs emerge that are better performing than the previous GATs. This represents a GAT that understood better that inherent randomness and in effect it provides more overall hits and also you may notice successive hits at that GAT's graph (red line or blue line going up). So, a GAT is not a different algorithm but more of a different means of evolving randomness inherent in the draws. If we have a full understanding of randomess in a GAT table, it is possible this GAT to constantly hit the winning combination. This is extreme scenario really but observing one GAT table even hitting once the winning combination with the bare minimum of numbers over 100 tested draws it means at some point the randomness understanding evolved in such a way that it could provide that hit. Full randomess understanding might not be possible but nevertheless, the more GATs analysed the more hits production will emerge.
Someone could say, why not use normal AI - neural networks processes. Well, the answer is simple to that. AI needs define patterns to be observed in order to make sense, similar to an OCR program that uses AI - neural networks. An A will always be an A even if we draw the left side of that a like a curve; the lines (patterns) defining that letter are obvious.
In lotto draws we don't have such apparent patterns and even if one exists, is very short-living which makes it almost impossible to detect and utilize. Therefore AI techiques are not really good for predicting lotto draws because we really don't have obvious patterns there to take advantage of. On the other hand, GAT attempts to evolve the inherent randomness so the randomness itself is used to make a prediction. Each GAT actually represents a state of that randomness "understanding".
cheers
lottoarchitect

elmoo168
Casual
Casual
Posts: 16
Joined: Wed Dec 21, 2011 4:55 pm

Re: New GAT 2.2 coming soon!

Post by elmoo168 » Tue Feb 07, 2012 11:15 pm

Hi lottoarchitect,

Thanks for that explanation, it makes more sense now! I originally thought it was tied to the number of different combinations of the lottery, so for a 5/35 lotto I thought I didn't have to go above 324k GATs. I know now that's not the case.

Thanks again!

P.S. Any chance the registered users can get a beta of the program so we can try it out?!? :)

User avatar
lottoarchitect
Site Admin
Posts: 1534
Joined: Tue Jan 15, 2008 5:03 pm
Location: Greece
Contact:

Re: New GAT 2.2 coming soon!

Post by lottoarchitect » Wed Feb 08, 2012 3:20 pm

A pre-release version of the new 2.2 engine has been made available at the account management. Just extract the zip file at the current installation folder. It is just the exe file so make a shortcut of it to your desktop for quick access.

A brief explanation of the changes at the options panel (since no help file is available yet):
- we don't have anymore the maximum GATs selection since now the engine runs for as long as we let it to.
- Added a few speed-up parameters: Minimum/maximum req. numbers, minimum/maximum category and "modes to compute" . Use these to speedup the computation. A value of 0 means "compute everything" so if we set the Minimum/maximum req. numbers to "5-0" this means we'll keep track of GATs performance at the range of 5-maximum req. numbers. If you don't pick numbers e.g. beyond 12 or not interested in partial numbers selection (e.g. lower than 5 picked numbers) then set the range to 5-12 to speedup the process and reduce memory demands. The same applies to the hits category range. Finally if you are not interested in the "absolute mode", just enable the "augmentative only".
- "GATs to retain". This is important when utilizing the run factor. Since the engine now discards any GAT table that hasn't performed best over the tested draws, if we want to keep using a particular GAT in the future (with the run factor), this GAT might be rejected by the process (since other GATs might emerge as better performing). Any GAT number entered in that box will be retained, regardless if the engine indicates it should be rejected. To access those GATs at the "GAT table" view, just click the GAT ID label end enter its number.
- No more calibration and progress bar indication. In order to offer full speed to the engine, now we display the currently processed GAT ID only. This value is updated only when a change has occurred at the currently displayed panorama. So after a while the engine might seem doing nothing since fewer and fewer GATs will emerge as better (so to update the panorama display). To check which GAT is currently processed, just move the pointer over the text.
- The predicted numbers display which colored any correct number with a green color does not work in that version since there isn't anymore this gradual stepping over the tested draws. Any GAT shown at the "GAT table" panel is fully computed and the predicted numbers shown are for the next draw to come.

Since this is a beta release, probably a menu item will be added to quickly select those GATs and other additions/changes might occur till the final release and an option to keep the settings used at the options panel automatically. You may also comment for additions/other functionality before this release is finalized.


cheers
lottoarchitect

spot9
Casual
Casual
Posts: 10
Joined: Thu Jul 03, 2008 1:45 am

Re: New GAT 2.2 coming soon!

Post by spot9 » Wed Feb 08, 2012 8:54 pm

lottoarchitect wrote:A pre-release version of the new 2.2 engine has been made available at the account management. Just extract the zip file at the current installation folder. It is just the exe file so make a shortcut of it to your desktop for quick access.
Hi Anastasios,
I Tried the above, but now neither 2,1 or 2.2 will work. The message from the server is that I will have to reset my account!
spot9

User avatar
lottoarchitect
Site Admin
Posts: 1534
Joined: Tue Jan 15, 2008 5:03 pm
Location: Greece
Contact:

Re: New GAT 2.2 coming soon!

Post by lottoarchitect » Wed Feb 08, 2012 9:36 pm

This will occur if you place the downloaded exe file in a different folder. It must be in the same folder current GAT is. Also, if you try to run both programs, even if they reside in the same folder, make sure to let each instance fully open. If you quickly attempt to open another instance of the program this problem will also occur (see the help file - troubleshooting).

lottoarchitect

spot9
Casual
Casual
Posts: 10
Joined: Thu Jul 03, 2008 1:45 am

Re: New GAT 2.2 coming soon!

Post by spot9 » Wed Feb 08, 2012 10:01 pm

lottoarchitect wrote: It must be in the same folder current GAT is in.
lottoarchitect
Hi Anastasios,
I did copy it into the same folder.
Can not get either program out of demo mode so
it seems I have no choice but to reset the account.
spot9

User avatar
lottoarchitect
Site Admin
Posts: 1534
Joined: Tue Jan 15, 2008 5:03 pm
Location: Greece
Contact:

Re: New GAT 2.2 coming soon!

Post by lottoarchitect » Wed Feb 08, 2012 10:13 pm

You have to make an account reset, once a lock occurs it can't be unlocked. The above mentioned are reasons to get in an account lock. The problem you faced which locked the account is probably you tried to open quickly the program twice without letting the first instance fully open. I'll try to solve this issue so an account lock will not occur in that case. Don't forget to set the compatibility to the same settings used to run current GAT (if you use Vista/windows 7).

lottoarchitect

spot9
Casual
Casual
Posts: 10
Joined: Thu Jul 03, 2008 1:45 am

Re: New GAT 2.2 coming soon!

Post by spot9 » Wed Feb 08, 2012 10:17 pm

Thanks very much for your help Anastasios.
spot9

Post Reply

Who is online

Users browsing this forum: No registered users and 1 guest